Distributed Coordination for Autonomous Guided Vehicles in Multi-agent Systems with Shared Resources
The decentralized path planning technique proposed in this thesis solves major challenges in the domain of MAS. These challenges are trajectory planning and collision avoidance. Generally, in a shared infrastructure where several agents aim to use limited capacity resources, finding a set of optimal and conflict-free paths for each single agent is the most critical part. The purpose of this research is designing a decentralized framework to coordinate the behavior of a number of agents in dynamic environments where continual planning and scheduling are required. In this research, a set of tasks will be assigned to the agents. Based on the origin and the destination of each task, the optimal path will be selected while head-on collision between the agents are avoided. The proposed algorithm has been evaluated for different scenarios in simulated experiments. Numerical comparison and verification against centralized algorithm shows that decentralized algorithm removes deadlock generated in centralized system at the cost of increasing the travel time.